Skip to content

feat(mcp): record safe tool input names - #989

Open
gesh wants to merge 3 commits into
mainfrom
posthog/mcp-input-names
Open

gesh wants to merge 3 commits into
mainfrom
posthog/mcp-input-names

Conversation

@gesh

@gesh gesh commented Sep 29, 2026

Copy link
Copy Markdown
Member

💡 Motivation and Context

MCP server owners need to measure which input names agents use. This must not expose argument values or unknown field names.

This PR ports the input-name and alias contract from @posthog/mcp 0.19.0 and 0.20.0. It adds $mcp_input_keys, $mcp_input_aliases_used, and a helper for custom dispatchers.

💚 How did you test it?

  • uv run pytest -q posthog/test/mcp --timeout=30
  • uv run ruff check posthog/mcp posthog/test/mcp/test_tool_input.py posthog/test/mcp/test_features_m4.py
  • uv run mypy --no-site-packages --config-file mypy.ini posthog/mcp | uv run mypy-baseline filter
  • make public_api_check

📝 Checklist

  • I reviewed the submitted code.
  • I added tests to verify the changes.
  • I updated the docs if needed.
  • No breaking change or entry added to the changelog.

If releasing new changes

  • Ran sampo add to generate a changeset file

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Codex implemented the agreed Python port. The implementation keeps schemas isolated by session and redacts unknown names by default.


Created with PostHog Desktop

@gesh gesh self-assigned this Sep 29, 2026
@github-actions

github-actions Bot commented Sep 29, 2026 •

Copy link
Copy Markdown
Contributor

posthog-python Compliance Report

Date: 2026-10-01T16:07:50.736753+00:00
Duration: 260188ms

✅ All Tests Passed!

121/121 tests passed


Capture_V1 Tests

✅ 95/95 tests passed

View Details
Test Status Duration
Endpoint And Method.Targets V1 Endpoint ✅ 519ms
Endpoint And Method.Does Not Use Legacy Endpoints ✅ 512ms
Required Headers.Has Authorization Bearer Header ✅ 512ms
Required Headers.Has Content Type Json ✅ 513ms
Required Headers.Has Posthog Sdk Info Format ✅ 512ms
Required Headers.Has Posthog Attempt Header ✅ 512ms
Required Headers.Has Posthog Request Id ✅ 511ms
Required Headers.Has Posthog Request Timestamp ✅ 512ms
Required Headers.Has User Agent ✅ 511ms
Body Format.Body Has Created At And Batch ✅ 511ms
Body Format.No Api Key In Body ✅ 511ms
Body Format.No Sent At In Body ✅ 513ms
Event Format.Event Has Required Root Fields ✅ 510ms
Event Format.Event Uuid Is Valid ✅ 512ms
Event Format.Event Timestamp Is Rfc3339 ✅ 511ms
Event Format.Non Utc Event Timestamp Is Converted To Utc ✅ 520ms
Event Format.Distinct Id Is String ✅ 511ms
Event Format.Distinct Id At Root Not Properties ✅ 511ms
Event Format.Custom Properties Preserved ✅ 512ms
Event Format.Set Properties Preserved ✅ 511ms
Event Format.Set Once Properties Preserved ✅ 512ms
Event Format.Groups Properties Preserved ✅ 512ms
Event Format.Sdk Generates Uuid If Not Provided ✅ 512ms
Event Format.Event Has Required Root Fields Batch ✅ 515ms
Event Format.Event Uuid Is Valid Batch ✅ 516ms
Event Format.Event Timestamp Is Rfc3339 Batch ✅ 517ms
Event Format.Distinct Id Is String Batch ✅ 518ms
Event Format.Distinct Id At Root Not Properties Batch ✅ 516ms
Event Format.Custom Properties Preserved Batch ✅ 516ms
Event Format.Set Properties Preserved Batch ✅ 519ms
Event Format.Set Once Properties Preserved Batch ✅ 517ms
Event Format.Groups Properties Preserved Batch ✅ 516ms
Event Format.Sdk Generates Uuid If Not Provided Batch ✅ 516ms
Batch Behavior.Multiple Events In Single Batch ✅ 521ms
Batch Behavior.Batch Envelope Smoke ✅ 519ms
Batch Behavior.Flush With No Events Sends Nothing ✅ 507ms
Batch Behavior.Flush At Triggers Batch ✅ 1016ms
Batch Behavior.Created At Reflects Batch Creation Time ✅ 511ms
Deduplication.Generates Unique Uuids ✅ 521ms
Deduplication.Different Events Same Content Different Uuids ✅ 516ms
Deduplication.Preserves Uuid On Retry ✅ 6517ms
Deduplication.Preserves Timestamp On Retry ✅ 6521ms
Deduplication.Preserves Uuid And Timestamp On Batch Retry ✅ 6524ms
Deduplication.No Duplicate Events In Batch ✅ 523ms
Header Behavior On Retry.Attempt Header Starts At One ✅ 512ms
Header Behavior On Retry.Attempt Header Increments On Retry ✅ 13532ms
Header Behavior On Retry.Request Id Preserved On Retry ✅ 6522ms
Header Behavior On Retry.Different Requests Have Different Request Ids ✅ 3023ms
Header Behavior On Retry.Request Timestamp Changes On Retry ✅ 6521ms
Response Format Validation.Success Response Has Uuid Keyed Results ✅ 513ms
Response Format Validation.Success Response Has Ok For Each Event ✅ 518ms
Response Format Validation.Success No Retry After When All Ok ✅ 514ms
Response Format Validation.Success Retry After Present When Retry Events ✅ 1518ms
Response Format Validation.Success No Retry After When Drop Only ✅ 514ms
Response Format Validation.Response Echoes Request Id ✅ 511ms
Retry Behavior.Retries On 408 ✅ 6523ms
Retry Behavior.Retries On 500 ✅ 6518ms
Retry Behavior.Retries On 503 ✅ 8525ms
Retry Behavior.Retries On 504 ✅ 6522ms
Retry Behavior.Retryable Errors Have Retry After ✅ 3517ms
Retry Behavior.Respects Retry After On Retryable Error ✅ 11526ms
Retry Behavior.Does Not Retry On 400 ✅ 2514ms
Retry Behavior.Does Not Retry On 401 ✅ 2516ms
Retry Behavior.Does Not Retry On 402 ✅ 2515ms
Retry Behavior.Does Not Retry On 413 ✅ 2516ms
Retry Behavior.Does Not Retry On 415 ✅ 2516ms
Retry Behavior.Non Retryable Errors Have No Retry After ✅ 2519ms
Retry Behavior.Implements Backoff ✅ 22534ms
Retry Behavior.Max Retries Respected ✅ 22529ms
Partial Batch Handling.Handles 200 Full Success ✅ 2513ms
Partial Batch Handling.Handles 200 With All Ok ✅ 3524ms
Partial Batch Handling.Does Not Retry Dropped Events ✅ 3517ms
Partial Batch Handling.Does Not Retry Limited Events ✅ 3519ms
Partial Batch Handling.Prunes Ok Events On Partial Retry ✅ 6522ms
Partial Batch Handling.Prunes Dropped Events On Partial Retry ✅ 6524ms
Partial Batch Handling.Retries Only Retry Events From Partial ✅ 6527ms
Partial Batch Handling.Partial Retry Preserves Uuids ✅ 6521ms
Partial Batch Handling.Partial Retry Attempt Header Increments ✅ 6520ms
Partial Batch Handling.Partial Retry Request Id Preserved ✅ 6525ms
Partial Batch Handling.Respects Retry After On Partial ✅ 8522ms
Partial Batch Handling.Unknown Result Treated As Terminal ✅ 3518ms
Partial Batch Handling.Mixed Ok Drop Limited No Retry ✅ 3520ms
Compression.Sends Gzip Content Encoding ✅ 513ms
Compression.No Content Encoding When Disabled ✅ 511ms
Compression.Compressed Body Is Decompressible ✅ 512ms
Error Handling.Does Not Retry On Unknown 4Xx ✅ 2512ms
Event Options.Cookieless Mode Override ✅ 512ms
Event Options.Disable Skew Correction Override ✅ 512ms
Event Options.Process Person Profile Override ✅ 511ms
Event Options.Product Tour Id Override ✅ 511ms
Event Options.Unset Options Omitted ✅ 511ms
Event Options.Options Override In Batch ✅ 515ms
Geoip And Historical Migration.Geoip Disable Injected Into Properties ✅ 511ms
Geoip And Historical Migration.Historical Migration Set In Body ✅ 512ms
Geoip And Historical Migration.Historical Migration Absent By Default ✅ 511ms

Capture_Ai Tests

✅ 5/5 tests passed

View Details
Test Status Duration
Routing.Capture Ai Posts To Ai Endpoint ✅ 511ms
Routing.Capture Does Not Reroute Ai Named Events ✅ 512ms
Identity.Capture Ai Event Has Uuid ✅ 511ms
Identity.Capture Ai Keeps Supplied Uuid ✅ 512ms
Timestamp Format.Non Utc Event Timestamp Is Converted To Utc ✅ 513ms

Feature_Flags Tests

✅ 17/17 tests passed

View Details
Test Status Duration
Request Payload.Request With Person Properties Device Id ✅ 13ms
Request Payload.Flags Request Uses V2 Query Param ✅ 12ms
Request Payload.Flags Request Hits Flags Path Not Decide ✅ 12ms
Request Payload.Flags Request Omits Authorization Header ✅ 12ms
Request Payload.Token In Flags Body Matches Init ✅ 12ms
Request Payload.Groups Round Trip ✅ 13ms
Request Payload.Groups Default To Empty Object ✅ 12ms
Request Payload.Disable Geoip False Propagates As Geoip Disable False ✅ 12ms
Request Payload.Disable Geoip Omitted Defaults To False ✅ 12ms
Request Payload.Flag Keys To Evaluate Contains Only Requested Key ✅ 12ms
Request Lifecycle.No Flags Request On Init Alone ✅ 4ms
Request Lifecycle.No Flags Request On Normal Capture ✅ 511ms
Request Lifecycle.Two Flag Calls Produce Two Remote Requests ✅ 18ms
Request Lifecycle.Mock Response Value Is Returned To Caller ✅ 12ms
Retry Behavior.Retries Flags On 502 ✅ 316ms
Retry Behavior.Retries Flags On 504 ✅ 317ms
Side Effect Events.Get Feature Flag Captures Feature Flag Called Event ✅ 514ms

Feature_Flags_Local_Evaluation Tests

✅ 4/4 tests passed

View Details
Test Status Duration
Versioned Boolean Matching.Matching Version Missing ✅ 80ms
Versioned Boolean Matching.Matching Version 1 ✅ 77ms
Versioned Boolean Matching.Matching Version 2 ✅ 76ms
Versioned Boolean Matching.Version Only Reload 1 2 1 2 Missing ✅ 46ms

@greptile-apps

greptile-apps Bot commented Sep 29, 2026 •

Copy link
Copy Markdown
Contributor

Retrigger

[Medium risk] Adds safe input field name recording to MCP tool-call events.

The PR should not merge until declared input names are captured for referenced standalone FastMCP schemas.

Reviews (2) · Last reviewed commit: "fix(mcp): scope input schemas to tool ca..."

Comment thread posthog/mcp/_instrumentation.py Outdated
Comment thread posthog/mcp/_instrumentation.py Outdated
Comment thread posthog/mcp/_tool_input.py Outdated
Comment thread posthog/mcp/_tool_input.py Outdated
@gesh
gesh force-pushed the posthog/mcp-input-names branch from 2f602c9 to 57a3d9a Compare October 1, 2026 14:55
@gesh
gesh marked this pull request as ready for review October 1, 2026 15:07
@gesh
gesh requested a review from a team as a code owner October 1, 2026 15:07
@gesh
gesh requested a review from a team October 1, 2026 15:08
Comment thread posthog/mcp/_instrument_lowlevel.py Outdated
gesh added 3 commits October 1, 2026 19:02
Generated-By: PostHog Desktop
Task-Id: dda38523-5b8d-47b6-a11a-64cd59a812fe
Generated-By: PostHog Desktop
Task-Id: dda38523-5b8d-47b6-a11a-64cd59a812fe
@gesh
gesh force-pushed the posthog/mcp-input-names branch from c8b2409 to c520c54 Compare October 1, 2026 16:02

@lucasheriques lucasheriques left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The no-values design and the analytics-key filtering hold up. Approving, with a few parity questions, since the same property would carry different data in Python and JS.

assert properties["$mcp_input_keys"] == ["a", "b"]


async def test_lowlevel_does_not_reuse_a_listed_schema_for_input_names():

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

question: Raw low-level servers always record ["[redacted]"], even after a listing, and this test locks that in. JS uses listed schemas per session here. I'm guessing 57a3d9a followed the same reasoning as _standalone_injected_parameters (listings can differ across requests). For names, though, a stale schema only risks recording a name the server declared, not a private one. Was there another concern?

]


def _aliases_used(

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

question: This differs from JS describeAliasesUsed, which takes the first alias present and never calls should_record_input_key. Here a too-long alias falls through to a later one, and the rule runs on the canonical name (absent from the input, with declared=True). Same input, different $mcp_input_aliases_used per SDK. Match JS, or land the stricter rule there too? Matching JS also removes recording_decisions.

model_ours = data.tool_model_parameter_injected.get(name, model_injectable)
if _dispatch_can_differ(high_level):
model_ours = False
input_schema = None

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

question: With _dispatch_can_differ, any middleware (even FastMCP's LoggingMiddleware) redacts every name: ['a', 'b'] becomes ['[redacted]']. The v2 path handles middleware by resolving through server.list_tools() in the request. Could 1.x do the same for input names?

return declared, injectable


def _input_schema_view(

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

question: Reading the code, $ref dereferencing only runs on the mcp 1.x path, while v2 passes raw tool.parameters. Would a root-$ref schema redact everything under mcp 2.x? If so, moving it into _declared_properties would cover every adapter.

return properties if isinstance(properties, dict) else {}


def _tool_input_schema_v2(high_level: Any, name: str) -> Optional[Dict[str, Any]]:

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nits:

  • This and _tool_input_schema in fastmcp are identical. Pass the schema to start_tool_call_lifecycle in every adapter and drop the replace calls.
  • README: document should_record_input_key, the 20/64 limits, and that $mcp_parameters still carries raw names.
  • The helper tests could be one parametrize table, like JS's it.each.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants